merge: AgentProject membership — the junction is now load-bearing

Merges feat/agent-project-membership (task 166/152). An agent can finally join
many projects: membership routes, per-membership role and optional brain
override, chat resolving context from the membership, and CLI attach-agent /
detach-agent / get memberships. The legacy Agent.projectId fallback is kept for
this release as the schema comment promises.

Conflict resolved in main.ts: that branch added its RBAC case to
mapUrlToPermission inline, while this branch had extracted the function to
middleware/url-permission.ts so it could be tested. Kept the extraction and
ported the case into it — joining a team is `edit:agents:<name>`, not
`create:agents`, which is the authority to mint a whole agent.

Also fixes something the extraction makes possible. security.test.ts contained a
hand-copied reproduction of mapUrlToPermission, because main.ts self-invokes
main() and nothing in it was importable. A security test asserting against a
copy stops testing the system the moment the real one changes, and this copy had
already drifted. It now imports the real function; all 18 assertions still pass,
against the code that actually runs.

Local: build clean, workspace 2550 passed, lint unchanged at 794 errors.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4wNHWf7xSwnZCWpJcyv9p

.taskmaster/tasks/task_046.md

@@ -0,0 +1,59 @@
+# Task ID: 46
+
+**Title:** Implement LLM-based prompt selection service
+
+**Status:** done
+
+**Dependencies:** 45 ✓
+
+**Priority:** high
+
+**Description:** Create a service that uses the heavy LLM to intelligently select relevant prompts based on tags and the full prompt index, understanding synonyms and context.
+
+**Details:**
+
+1. Create `/src/mcplocal/src/services/llm-prompt-selector.service.ts`:
+   ```typescript
+   export class LlmPromptSelectorService {
+     constructor(
+       private llmClient: LlmClient,
+       private fastLlmClient: LlmClient | null,
+       private tagMatcher: TagMatcherService  // fallback
+     ) {}
+     
+     async selectPrompts(tags: string[], promptIndex: PromptIndex[]): Promise<{
+       selected: Array<{ name: string; reason: string }>;
+       priority10: PromptIndex[];  // Always included
+     }> {
+       // 1. Extract priority 10 prompts (always included)
+       // 2. Generate missing summaries using fast LLM if needed
+       // 3. Send to heavy LLM:
+       const prompt = `
+         Given these keywords: ${tags.join(', ')}
+         And this prompt index:
+         ${promptIndex.map(p => `- ${p.name}: ${p.summary}`).join('\n')}
+         
+         Select the most relevant prompts for someone working on tasks
+         related to these keywords. Consider synonyms and related concepts.
+         Return a ranked JSON array: [{name: string, reason: string}]
+       `;
+       // 4. Parse LLM response
+       // 5. On LLM error, fall back to tag matcher
+     }
+   }
+   ```
+
+2. Handle LLM timeouts and errors gracefully with fallback
+
+3. Validate LLM response format
+
+**Test Strategy:**
+
+- Unit test: Priority 10 prompts always returned regardless of LLM selection
+- Unit test: LLM called with correct prompt format (mock)
+- Unit test: LLM response parsed correctly
+- Unit test: Invalid LLM response falls back to tag matcher
+- Unit test: LLM timeout falls back to tag matcher
+- Unit test: Missing summaries trigger fast LLM generation
+- Unit test: No LLM available uses tag matcher directly
+- Integration test: End-to-end selection with mock LLM